14800-DvRappard

110 Chapter 7 MR Data Analysis For quantification of metabolite concentrations measured by MRSI, we used the signal intensity of water in the unsuppressed reference scans obtained with head coil and body coil, i.e. the voxel-wise ratio SI body /SI head multiplied by the transmitter amplitude of the body coil. 15-17 All spectra within the VOI were quantified using LCModel. 18 Metabolite concentrations (in mmol/L VOI) were reported for creatine and phosphocreatine (Cr), NAA (including contributions of N-acetyl-aspartylglutamate), Cho, Ins, glutamate (Glu), and lactate (Lac). Because estimation of glutamine (Gln) was less reliable in the majority of spectra, indicated by Cramer Rao lower bounds (CRLB) above 20%, 18 we also reported Glx, which is the sum of Glu and Gln. Although the low concentration of Lac in controls and some patients inherently led to large CRLB values as well, we included all estimated Lac concentrations because they were necessary for comparison with patients in whom Lac was clearly detectable. Lesions were outlined and quantified on FLAIR using the semi-automatic tool clusterize 19 and linearly registered to 3D T1 (Figure 1). Lesion masks in 3D T1 were filled with the signal intensity of normal-appearing WM (NAWM) to ensure correct segmentation into grey matter (GM), WM and cerebrospinal fluid (CSF) with SIENAX from FMRIB’s software library FSL. 20 Because the MRSI slab may include subcortical areas, these were segmented with FIRST, also part of FSL. 20 Extrapolation analysis combining the MRSI slab and partial volume estimates of CSF and tissue from segmentation resulted in concentrations for GM, overall WM, and NAWM and lesions separately. In this analysis concentrations in CSF are assumed negligible, and concentrations in mM indicate mmol/L tissue. Because of large variability between patients in lesion volume, our main analysis involved concentrations in overall WM (NAWM + lesions), unless otherwise indicated. Subcortical areas had no or only a minor contribution to the MRSI slab and were taken into account for determining the other tissue concentrations, but not further evaluated. The quantitative analyses were not blinded for patient characteristics, but apart from the semi-automatic lesion segmentation, all analyses were user-independent. Axial T2-weighted images were scored using the MLD-MRI score. 6